Extraordinary acquirers Andrey Golubov, Alfred Yawson and Huizhong Zhang* November 2013 Preliminary Version Abstract Firm fixed effects alone explain more of the variation in acquirer returns than all the firm and deal specific characteristics combined. An inter-quartile range of acquirer fixed effects is over six times the average acquirer return of about 1% in our sample. Acquirer returns persist over time, but mainly at the extremes. The attributes of the top management team fail to explain the fixed effect. Firm- specific heterogeneity in acquirer returns suggests that some organizations are extraordinary acquirers irrespective of the leadership at the top and the deal structures they choose. Implications for the M&A research are discussed. JEL classification: G14; G34 Keywords: Mergers and Acquisitions; Acquirer Returns; Performance Persistence; Fixed Effects
Transcript
Extraordinary acquirers
Andrey Golubov, Alfred Yawson and Huizhong Zhang*
November 2013
Preliminary Version
Abstract
Firm fixed effects alone explain more of the variation in acquirer returns than all the firm and
deal specific characteristics combined. An inter-quartile range of acquirer fixed effects is
over six times the average acquirer return of about 1% in our sample. Acquirer returns persist
over time, but mainly at the extremes. The attributes of the top management team fail to
explain the fixed effect. Firm-specific heterogeneity in acquirer returns suggests that some
organizations are extraordinary acquirers irrespective of the leadership at the top and the deal
structures they choose. Implications for the M&A research are discussed.
JEL classification: G14; G34
Keywords: Mergers and Acquisitions; Acquirer Returns; Performance Persistence; Fixed
Effects
*Golubov is from Cass Business School, City University London ([email protected]). Yawson is from University of Adelaide Business School ([email protected]). Zhang is from University of Adelaide Business School ([email protected]). We thank Hendrik Bessembinder, Ettore Croci, Cláudia Custódio, B. Espen Eckbo, Alex Edmans, Eliezer Fich, John Forker, Sudipto Dasgupta, Jarrad Harford, Marcin Kacperczyk, Ronald Masulis, Holger Mueller, Tu Nguyen, Micah Officer, Takeshi Yamada, as well as seminar participants at City University London, University of Adelaide, University of Essex, University of Exeter, University of Sussex, University of Bath, University of Surrey, Queen’s University Belfast, FMA 2013 Annual Conference, NTU 2012 International Conference on Finance and Australasian Finance and Banking 2012 Conference for helpful comments and suggestions. Part of this work was conducted while Yawson was a Visiting Research Scholar at Cass Business School, City University London. All remaining errors are our own.
“Finally, knowledge of the source of takeover gains still eludes us” (Jensen and Ruback, 1983, p.47)
Thirty years after the seminal work by Jensen and Ruback (1983) their conclusion as to
the elusive nature of takeover gains is relevant as ever. Despite the explosion in large sample
studies of mergers and acquisitions (M&A) over the last three decades, the variation in
returns to acquisition activity has not been explained in any major way. For example, a
widely cited study by Moeller, Schlingemann and Stulz (2004) examines over 12,000 M&A
deals, and, employing an extensive list of determinants, is able to explain just over 5% of the
variation in acquirer returns (as judged by the Adjusted R2 of their main regression models).1
Similar, albeit smaller sample studies such as Masulis, Wang and Xie (2007) and Harford,
Humphery-Jenner and Powell (2012) report comparably low explanatory powers. If an
exhaustive list of factors in combination does not explain the variation in acquirer returns in a
systematic way, then what does?
In this paper we show that acquirer returns are best explained by an unobserved, time-
invariant, firm-specific factor. In line with prior research, we show that the explanatory
power of a comprehensive regression specification employing most of the widely used firm
and deal specific characteristics explains only 5.0-6.4% (Adjusted R2, 3.6-6.0%) of the
variation in acquirer returns. However, the same regression model augmented with acquirer
fixed effects explains almost half of the variation in acquirer returns. These findings suggest
that the source of acquirer gains is not deal but rather firm-specific. That is, some firms are
stellar acquirers irrespective of their time-varying attributes and the deal structures they
choose. The economic magnitude of the fixed effect is staggering. An inter-quartile range in
1 This is not to detract from the contribution of the Moeller et al. (2004) study. In fact, the acquirer size effect
they document has proven to be one of the most robust determinants of acquirer returns; we also confirm it in
this paper.
2
acquirer fixed effects is over six times the unconditional acquirer CAR of about 1% in our
sample.
We further show that acquirer returns are persistent over time. Acquirers in the top
performance quintile continue to make better acquisitions than acquirers from the bottom
performance quintile at least up to five years down the road. We find that persistence in
acquirer returns is concentrated mainly in the best performers. There is a positive dependence
of future returns on past returns at 80th percentiles of the return distribution, but not at the
mean.
A further intriguing question is what explains the acquirer fixed effect? Two plausible
candidates emerge. First, it could be the slow-moving attributes of the acquirer’s management
that are at play. For instance, it is possible that the effect is driven by the top management of
the firm. We probe into this “managerial” explanation and attempt to explain the acquirer
fixed effects using the attributes of the top management team found to be important by the
management science literature. We find that the inclusion of these variables does not improve
the explanatory power of the acquirer returns model, and does not detract from the economic
or statistical significance of the fixed effects. This suggests that it is the time-invariant
specifics of the firm, and not its management team that make acquirers extraordinarily good
or bad.2 It appears that acquisition skill (or lack thereof) is “wired into” the organizational
structure of firms, such that some of them are great acquirers irrespective of the leadership at
the top and the particular deal structures they choose. Finally we find a great deal of overlap
in the industry distribution of extraordinarily good and bad acquirers, further reinforcing the
idea that the sources of superior takeover performance are to be found within the firm and not
2 Another possibility is that the firm fixed effect stands in for the CEO effect, given that turnover events are rare.
We are not able to explicitly test for this due to extreme paucity of CEO moves between acquirers which are
required for separate identification of firm and CEO fixed effects.
3
within its environment. We conclude that some form of organizational knowledge is
responsible for the observed fixed effect and persistence in acquirer returns.
Our paper is related to several strands of literature. First, it is related to the growing
“fixed effects” literature. Bertrand and Schoar (2003) pioneered this line of research by
showing that manager fixed effects explain various corporate policies ranging from financing
to payout.3 Lemmon, Roberts and Zender (2008) show that capital structures are to a large
extent explained by time-invariant firm-specific attributes. Graham, Li and Qiu (2012)
document that manager-firm fixed effects explain a significant portion of the variation in
managerial compensation. In the M&A setting, Bao and Edmans (2011) show that a
significant advisor fixed effect in acquirer gains exists, and Kaplan and Schoar (2005) find
private equity house returns to be persistent. Second, our study is related to a strand of the
literature that attempts to establish whether “managers matter” in corporate decisions. Chang,
Dasgupta and Hilary (2010) provide evidence consistent with managerial effects explaining
corporate performance. In the M&A context, Custódio and Metzger (2013) show that
acquiring firm manager’s expertise in the target industry leads to better performance in
diversifying acquisitions. Finally, it is related to the M&A literature that attempts to explain
the distribution of acquirer returns. Notable examples are Moeller, Schlingemann and Stulz
(2005) and Fich, Nguyen and Officer (2013), who examine large M&A losses and gains,
respectively. We contribute to this literature by exploring the fixed effects and the attendant
persistence in acquirer returns more broadly. Our findings imply that, despite the perceived
saturation of the finance literature with M&A returns studies, we still appear to be missing
the major part of the puzzle and that the elusive driver of takeover gains is to be found within
3 More recently, Fee, Hadlock and Pierce (2013) question whether the manager-specific effects can be
interpreted as causal, and also raise a methodological issue regarding the use of the standard F-tests in assessing
the joint significance of the estimated fixed effects. We address this issue below.
4
the firm. We discuss the implications of our results for this literature and make suggestions
for further research.
The remainder of the paper is organized as follows. Section I presents our data and
establishes a highly significant firm fixed effect in acquirer returns. Section II confirms the
persistence of acquirer returns over time. We examine the sources behind the acquirer fixed
effect in Section III. Section IV discusses the key results and their implications. Finally, we
close in Section V with some concluding remarks and suggestions for further study.
I. Sample and Preliminary Results
I.A Sample Selection
The M&A data are sourced from the Thomson Financial SDC Platinum US M&A
database over the period from January 1, 1990 to December 31, 2011. We follow Fuller,
Netter and Stegemoller (2002) and Masulis et al. (2007) and impose the following
restrictions:
1. The bidder must be a US publicly listed company, and the target must be a US public,
private, or subsidiary firm.
2. The acquisition must be completed.
3. The acquirer must own less than 50% of the target stock before the acquisition and
achieve 100% after.
4. The transaction must be at least 1% of the acquirer’s market capitalization 11 days
before the announcement and also exceeds $1 million.
5. The bidder’s stock price data for 300 trading days prior to the announcement are
available from CRSP, and accounting data for the year-end immediately prior to the
announcement are available from Compustat.
6. Multiple deals announced by the same firm on the same day are excluded.
5
These requirements result in a sample of 12,491 transactions involving 4,128 unique
firms. We use the standard event study methodology to compute the cumulative abnormal
returns (CARs) of the sample acquirers over the event window (-2, +2) around the
announcement date.4 The CARs are measured as returns in excess of those predicted by the
market model with a benchmark being the CRSP value-weighted index and parameters
estimated over a period from 300 to 91 days prior to the announcement. In the empirical tests
that follow we work with three samples. The first sample consists of all 12,491 deals
identified above. There are 4,128 unique acquirers in the full sample. The other two samples
comprise only frequent acquirers to enable us test persistence in returns. Our first definition
of frequent acquirers (FNS serial acquirers hereafter) follows Fuller et al. (2002), which
requires at least 5 deals to be completed by the same acquirer within a 3-year period. This
definition reduces our sample quite dramatically and leaves us with 2,611 deals made by 333
unique acquirers. As an intermediate sample, we also define frequent acquirers as those
which complete at least 2 deals in any 3-year period (alternative serial acquirers hereafter).
Under this alternative definition we obtain 9,373 deals conducted by 2,219 unique acquirers.
Samples similar to ours have been extensively used in previous studies, so we refrain
from presenting elaborate descriptive statistics but verify that they are in line with prior
studies such as Masulis et al. (2007), Golubov, Petmezas and Travlos (2012) and Harford et
al. (2012). However, one noteworthy observation emerges. Restricting the sample to serial
acquirers defined as those having completed at least 2 acquisitions in a 3-year period
(alternative serial acquirers) reduces the sample by less than 25%, from 12,491 to 9,373
deals. That is, almost all deals are done by frequent acquirers and there is virtually no such
thing as a one-off deal in a typical M&A sample found in most papers. The implication of
this is two-fold. First, this structure of the data lends them well to the fixed effects analysis
4 We also used CARs over event windows (-1, +1) and (-5, +5), as well as market-adjusted returns instead of the
market model. All results remain unchanged.
6
that we undertake below. Second, it suggests that M&A studies employing long-run abnormal
stock returns or operating performance improvements (typically measured over 3 years
following the deal) stand little chance of attributing the results to a particular deal (or deal
characteristic) unless they exclude all frequent acquirers, which leaves only a small and, most
likely, selected and unrepresentative sample.
I.B First Results
The first part of our empirical analysis is to a large extent modelled along the work of
Bao and Edmans (2011) who examine investment bank fixed effects in M&A returns. We
begin with a cross-sectional regression of acquirer CARs for the 3 samples to serve as our
benchmark specification. We employ an extensive list of explanatory variables found in most
recent and influential acquirer returns studies. Specifically, we follow Masulis et al. (2007),
Golubov et al. (2012) and Harford et al. (2012) and control for acquirer size, Tobin’s Q, stock
price run-up, idiosyncratic stock return volatility (sigma), free cash flow and leverage. We
also include deal specific controls, namely, relative size, industry relatedness of the target,
tender offer and hostile dummies and a set of interactions between target listing status and the
method of payment. We report our first results in Table 1. All variables are defined in
Appendix A.
[Please Insert Table 1 Here]
Most of the estimated coefficients are of the expected signs and consistent with prior
studies although not always statistically significant. The most significant variables across all
the three regressions are acquirer size and the interaction term of public targets and stock
payment, which are both negatively associated with acquirer CARs. Tobin’s Q and stock
price run-up are consistently negative but significant only in the full and the alternative serial
acquirer samples. The interaction term between public target and all cash deals is negatively
7
associated with CARs even though it is significant in the serial acquirer sub-samples only.
Further we find a positive effect of sigma and relative size across all the three models
although not always significant.
Most importantly, the R2 (adjusted R2) of these regressions are very modest, but
comparable to those in prior studies (e.g. Moeller et al. (2004), Masulis et al. (2007), Harford
et al. (2012)). All of the variables combined explain only 5-6% of the variation in acquirer
returns. In short, a comprehensive regression model fails to capture the variation in acquirer
returns in any major way. We now compare these results to a simple fixed effects model
reported in Table 2.
[Please Insert Table 2 Here]
We first focus our discussion on Panel A which employs the full sample, and then
comment on the results for the frequent acquirer subsamples. Similar to Bertrand and Schoar
(2003), we report F-statistics for test of the joint significance of the different sets of acquirer
fixed effects for each of the three samples. Strikingly, a simple model with an acquirer fixed
effect reported in the first row produces an R2 (adjusted R2) of 46% (19.4%). The fixed effects
are highly jointly significant as evidenced by the F-statistic. Moving from the first row to the
fourth, we first add the year fixed effects, followed by deal characteristics and then by time-
varying firm level control variables to the basic fixed effects model. We find that the
inclusion of these additional variables contributes only modestly to the explanatory power of
the basic model. Specifically, the R2 (adjusted R2) increases by only 2.7% (3.7%) as we move
from the first row to the fourth. Moreover, the acquirer fixed effects remain highly
statistically significant: the F-tests in all cases are significant at the one percent level leading
us to reject the null hypothesis of no significant joint effects.
We replicate the fixed effect tests for the two frequent acquirer subsamples and report
the results in Panels B and C. In these subsamples each acquirer is found strictly more than
8
once, and hence the firm fixed effects should be more precisely estimated. In all cases we
find that the firm fixed effects are highly statistically significant (at the one percent level)
with the only exception of model (3) in the FNS serial acquirers subsample. The R2 (adjusted
R2) of the acquirer fixed effect models in the first rows of Panel B and C are 29.8% (8.0%)
and 15.0% (2.6%), respectively. Moving from the first to the fourth row by adding year fixed
effect, deal and acquirer characteristics increases the R2 (adjusted R2) by 3.3% (3.9%) and
4.7% (4.0%), respectively. While the R2 (adjusted R2) are more modest in the frequent
acquirer subsamples, this, in fact, is further evidence in favor of the fixed effects story. The
decrease in explanatory power is expected for the following reason. In Panel A, the ratio of
unique firms to total number of acquisitions is 0.33. In Panel B, the ratio is 0.24, and in Panel
C the ratio is 0.13, as the serial acquirer criterion becomes more stringent. If it is the firm
effect that explains acquirer gains, then the more of these firm effects there are in the model,
the better the fit. Consequently, as the number of firms relative to the sample size declines, so
does the explanatory power of the model. Therefore, the fact that the R2 goes down as we
move down the Panels further reinforces the idea that variation in acquirer returns is firm
specific rather than deal specific. Second, even these more modest R2 are higher than those in
our benchmark regressions in Table 1. Thus, fixed effects alone explain more of the variation
in acquirer returns than many of the important variables identified by prior literature
combined.
So far we have established the statistical significance of acquirer fixed effects and to
reinforce the substance of these results, we evaluate the economic magnitude of the fixed
effects. In Table 3 we report the inter-quartile ranges of the estimated fixed effects for the 3
samples and 4 specifications ranging from fixed effects only to year fixed effects and a full
set of deal and acquirer characteristics in that order.
[Please Insert Table 3 Here]
9
As reported in Panel A, the inter-quartile range of the returns for the full sample is
between 6.1% and 7.8%. When compared to the unconditional mean acquirer CAR of 1.16%
in the full sample, these results are staggering. Firm fixed effects are on average over six
times the average acquirer CAR. There is a wide gap in returns between very good and very
bad acquirers. Note that the model in row (4) includes a full set of controls, so that the
estimated acquirer fixed effects are orthogonal to deal-specific features and time-varying
firm-specific characteristics. Our results show that acquirers are either extraordinarily good or
bad irrespective of the deal structures they choose.
Panels B and C demonstrate the economic significance of acquirer fixed effects in the
serial acquirer subsamples. Depending on the set of controls, the inter-quartile range is
between 4.85% and 6.33% for the alternative serial acquirers subsample, and between 3.47%
and 6.25% for the FNS serial acquirer sample. When year fixed effects, deal and time-
varying firm specific characteristics are included in row four, the difference between the 25 th
and 75th percentile is 6.33% and 6.25% for the alternative serial acquirers and the FNS serial
acquirers, respectively. Again, on the backdrop of the unconditional mean acquirer CARs of
0.97% and 0.54% in the two subsamples, these effects are too large to be ignored. Thus, the
economic significance of the frequent acquirer results confirms that certain acquirers are
systematically associated with extraordinary acquisition performance.
I.C Robustness of the Acquirer Fixed Effect
Recently, Fee, Hadlock and Pierce (2013) criticize the use of standard F-test
procedures in establishing the joint significance of the estimated fixed effects. Replicating the
analysis of Bertrand and Schoar (2003), they show that after scrambling the data and
randomly assigning CEOs to firms – thereby destroying any CEO effect in the data – the
standard F-tests and the associated p-values are hardly affected. This indicates a strong fixed
effect even when none is present in the data by construction, casting doubt on the validity of
10
inferences based on standard F-tests in this context. We take these concerns seriously and
perform a similar data scrambling exercise to establish robustness of our main result.
Specifically, we follow Fee, Hadlock and Pierce (2013) and break the structure of the
data by randomly allocating deals to firms. We perform such Monte Carlo permutations of
the data 1000 times, each time re-estimating the fixed effects models in Table 2 and recording
the F-test on the joint significance of the firm fixed effect. If the firm-specific effect in
acquirer returns is genuine, we would expect it to disappear when the deals are randomly
allocated to firms. Table 4 reports the results of this placebo-type analysis.
[Please Insert Table 4 Here]
Reported in the table are the median F-tests and associated p-values from the 1000
random permutations of the data (the results based on the mean are identical). In all cases we
find that the F-test loses statistical significance and does not reject the null of no significant
firm-specific effects in acquirer returns. This is relieving, as no firm effect is in fact present in
the data following the permutations. It turns out that, in our context, the standard F-test
performs well, identifying a significant fixed effect when it appears to be present and failing
to identify one when there is none by construction. Our subsequent tests of the persistence in
acquirer returns further alleviate the concerns that the fixed effect is spurious.
II. Persistence of Acquirer Returns
The presence of a strong acquirer fixed effects implies that acquirer returns are
persistent over time. In this section we perform formal tests of persistence in acquirer returns.
Persistence tests explicitly require multiple acquisitions by all acquirers over time and as a
consequence we restrict this part of the analysis to the two serial acquirer subsamples. Our
methodology here is similar to Jegadeesh and Titman (1993) for stocks, Carhart (1997) for
mutual funds, and Bao and Edmans (2011) for investment bank advisors. We sort serial
11
acquirers into quintiles based on their average CARs over the last 3-year period (RET)
consistent with our definitions. For each quintile, we compute the average RET to the future
acquisitions made by all acquirers within that quintile over the next k calendar years, where k
= (1, 2, 3, 4, 5). We then test for the difference in means between the top (Q5) and the bottom
(Q1) quintiles.
[Please Insert Table 5 Here]
Table 5, Panel A reports persistence in raw CARs and the results are consistent across
the two subsamples. In both subsamples, the differences in CARs between Q5 and Q1 are
positive and statistically significant from k = 2 to k = 5. The lack of significant results for k=1
can be attributed to the small number of deals that are conducted within one year. We repeat
the persistence tests for residual CARs (RETRES) defined as the average residual CAR
obtained from Table 1 and report the results in Panel B. This is to ensure that the persistence
in raw acquirer returns is not driven by firm or deal specific characteristics (RETRES is
orthogonal to them). For both serial acquirer subsamples, there is persistence in residual
CARs and the pattern is consistent with those reported for the raw CARs.
The persistence tests above for both RET and RETRES are based on equally-weighted
average acquirer returns. However, such persistence may be misleading if an acquirer is good
at conducting relatively small deals, but destroys a lot of value when it comes to large
acquisitions. To rule the possibility of relatively small deals driving the averages out, we
perform value-weighted persistence tests. Panel C reports persistence in transaction-value
weighted RET and RETRES, where the weight is the ratio of the deal value to the sum of
transaction values of that acquirer over the period in which the performance is measured. The
value weighting does not alter the persistence results. We find persistence in both serial
acquirer subsamples and the differences between Q5 and Q1 are statistically and
economically significant in four out of five cells in each subsample. The results are broadly
12
consistent when transaction-value weighted RETRES is used in the persistence test. Besides,
RETRES is by construction orthogonal to deal-characteristics, including the relative size of
the deal.
It is remarkable to observe that the returns to the lower quintiles in each of the
significant results are negative whereas they are positive for the upper quintile. These results
provide additional evidence that certain acquirers are indeed extraordinary as they
persistently generate positive average returns in their acquisitions.
We also perform multivariate versions of the persistence tests. Multivariate
regressions allow us to use all firms and not just those in the top and bottom quintiles and
thus establish whether the persistence is characteristic of all acquirers or just those at the
extremes. Panel A of Table 6 reports the results of a simple OLS regression of future returns
(measured over 1, 2, 3, 4 and 5 years) on past returns (measured over 3 years consistent with
our definitions). Interestingly, this specification reveals no significant association between
future and past returns of the same acquirer, apart from the last two columns where the future
return is measured over 4 and 5 years - but the statistical significance is marginal (10% level).
This suggests that persistence in returns as documented in Table 5 is indeed concentrated at
the extremes, and there is little-to-no dependence of future returns on past returns on average.
Motivated by the significant univariate differences between the best and the worst
acquirers and the lack of a strong significant association between future and past returns on
average, we perform quantile regression analysis to further explore the persistence
phenomenon. Whereas an OLS regression estimates the conditional mean function, quantile
regressions allow for the estimation of the conditional n th percentile of the distribution as a
function of the explanatory variable(s). The coefficients in a quantile regression are
interpreted as the effect of a one unit change in the explanatory variable on the n th percentile
of the dependent variable. We model the 20th and the 80th percentiles of the future returns
13
distribution as a function of past returns of the same acquirer and report the results in Panels
B and C of Table 6.5 The 20th and 80th percentiles are approximately consistent with the
quintile results.
The results reported in Panel B show that past return is unable to explain future returns
in the quantile regressions estimated at 20th percentile point. Past RET is consistently
insignificant in all the estimated models with the exception of k = 5 where it is positive and
significant in the FNS sample. However, Panel C reveals a different picture. We find that
there is a strong positive association (all coefficients are significant at the 1% level) between
past and future returns at the 80th percentile point of the future returns distribution, and the
results are highly robust across both definitions of serial acquirers and across time horizons
over which future returns are measured. The intercepts are positive and significant at the top
(the 80th percentile) and negative and significant at the bottom (the 20th percentile) of the
distribution, consistent with our setup. These results continue to hold when RETRES instead
of the raw returns is used in the regression tests of persistence, with the exception that some
of the 20th percentile quantile regression specifications also produce significant associations
between future and past returns (reported in Appendix B).
[Please Insert Table 6 Here]
III. What Explains the Fixed Effect?
Having established a significant acquirer fixed effect and its flip side, persistence in
acquirer returns, we turn our attention to the potential sources of the fixed effects. In this
setting, we attempt to explain the economic force behind the statistical concept of fixed
effects and persistence in returns. In principle, the force behind the fixed effect could be
attributed to the uniqueness of the firm (i.e. its tangible or intangible assets, or the
5 We also experiment with 10th and 90th percentiles, but do not find consistently significant results.
14
processes/organizational knowledge) or special talents possessed by its management. Existing
literature stresses the role of CEOs in firm performance viewing them as the sole executives
in charge of core corporate development activities such as acquisitions (e.g., Roll (1986),
Hartzell, Ofek and Yermack (2004), Billett and Qian (2008), Aktas, de Bodt and Roll (2009),
Aktas, de Bodt and Roll (2011), Aktas, de Bodt and Roll (2013)). Since CEO turnover events
are rare for most firms, it is possible that the firm effect is, in fact, the CEO effect. In order to
disentangle the firm effect from the CEO effect one needs to estimate a model with both firm
and CEO fixed effects. Naturally, for the CEO fixed effect to be identified separately from
the firm effect one needs to observe a given CEO conducting deals in at least two different
firms. Thus, not only a sample of CEO moves is required, but these moves have to be
between acquiring firms in our sample. We are able to identify only 110 deals conducted by
CEOs who can be found in at least two different acquiring firms in our full sample, with 60
and 0 observations in the alternative and FNS serial acquirer samples, respectively. This data
limitation precludes any meaningful analysis in this regard. This is despite our best efforts to
supplement the standard CEO data from Compustat’s Execucomp – the usual source of data
on corporate executives – with that from BoardEx whose coverage is broader.6 Therefore, we
cannot rule out that the observed firm fixed effect picks up the CEO effect.
Alternatively, the firm fixed effect could be picking up slow-moving attributes of the
managerial team, and one can look at the latter in order to establish whether they on their own
affect acquirer returns. Besides, a singular focus on the CEO as the sole driver of acquisition
decisions may be to narrow as it overlooks the interdependence among key executives in
organizational structures. In fact, practitioners often emphasize the importance of managerial
teams in making M&A deals a success or a failure. We therefore probe further into the
managerial explanation for the observed firm fixed effects in acquirer returns by borrowing
6 We would like to thank Cláudia Custódio and Daniel Metzger for kindly sharing their extended Execucomp-
BoardEx CEO dataset used in Custódio and Metzger (2013).
15
from the management science literature and identifying a set of variables characterizing the
managerial team as a whole that has been found to affect various corporate outcomes.
The data on the top management team is extracted from Compustat’s ExecuComp
database.7 We define the top management team as executives with a listed title above the
vice-president level reported in ExecuComp as they constitute executives at the senior-most
level. This is consistent with Chemmanur and Paeglis (2005) and Hambrick, Cho and Chen
(1996), among others. We take into account the dynamics in top management team over time
by measuring all the variables at the end of the most recent fiscal year prior to the
announcement date. We use several variables to capture different dimensions of the top
management team talent across firms.
Our first measure is team size. The numerical strength of the team reflects the managerial
resources available to the firm for it brings diversity to corporate decision making in areas
such opportunity seeking and negotiations. Compared with small-sized teams, for example,
large top management teams are able to enjoy a broader range of perspectives on a greater
number of items, critical judgments and alternative solutions for conducting comprehensive
search and analysis of strategic options (Haleblian and Finikelstein (1993)). Such increased
resources and capabilities can result in high-quality acquisition decisions and superior
performance. Large top teams are, however, prone to conflicts and cooperation problems that
would otherwise be absent in small groups (Jehn (1995)). Nevertheless, the complex, non-
routine nature of M&As makes it possible that the benefits of enhanced capabilities accruing
7 ExecuComp collects up to 9 executives from each company’s annual proxy statement (SEC form DEF14A) for
a given year, and hence, may not capture all of the company’s managers but for the purpose of our analysis this
information is sufficient. According to the SEC DEF 14A filling rules, a company is required to fully disclose
information about compensations received by its most senior executives and directors. The executive officers
named in a proxy statement are the most influential executives in the corporate decision making process and
should wield the greatest impact on acquisition strategies.
16
to large-sized teams would outweigh the costs associated with coordination problems
(Haleblian and Finikelstein (1993)).
Another widely used indicator of top management team capability is team tenure. Prior
evidence suggests that top team tenure is associated with persistence in strategic direction
(Finkelstein and Hambrick (1990)). Higher average tenure can indicate greater cohesion and
shared experiences in strategic decision making. Consistent with Chemmanur and Paeglis
(2005), top team tenure is calculated as the average number of years top team members have
worked in the acquiring firm.
Long tenure, however, may create increasing rigidity and complacency in a team’s
interaction process. It is therefore critical for long-tenured teams to possess certain degrees of
heterogeneity to offer new information sources and introduce new perspectives into the
decision making. While disagreements are more likely to be present in heterogeneous teams,
resolving such disagreements encourages team members to think carefully about the
appropriateness of the proposed strategic solution. This is likely to initiate extensive
investigations necessary for uncovering errors and producing sound evaluation results and
corporate decisions (Miller, Burke and Glick (1998)). In support of this view, prior studies
show a positive link between top team heterogeneity and firm performance, suggesting that
cognitive diversity is a valuable resource to a firm. Hambrick et al. (1996) for example, find
that top management teams with greater tenure heterogeneity enjoy higher growth rates in
both market share and profits. We define heterogeneity in team tenure, as the coefficient of
variation in team tenure.
The average age of the top management team is used as an additional proxy for the
general experience of the top team members having worked within and outside the acquiring
firm.
17
Finally, we consider the effect of a powerful CEO who can potentially make important
corporate decision on a stand-alone basis, disregarding other top managers’ views. This may
diminish any efficiency gains from a team, as team members may feel reluctant to participate,
share information or report idea that run counter to the CEO. Following Finkelstein (1992),
Hambrick and D'Aveni (1992) and Hayward and Hambrick (1997), among others, CEO
power is measured using the pay differential between the CEO and other top managers,
defined as salary plus bonus in the most recent fiscal year prior to the announcement date
scaled by the average salary plus bonus of the other top management team members.
Our strategy is as follows. We augment the fixed effects regression model in line 4 of
Table 2 with the managerial talent variables. We first examine whether these variables are
significant determinants of acquirer returns. Results are reported in Panel A of Table 7.
Surprisingly, none of the managerial talent variables explain acquirer returns in the full
sample. We find, however, that tenure heterogeneity and average age are statistically
significant in the FNS subsample whereas average tenure is positive and significant in the
alternative subsample. Even though the managerial talent variables play a role in the frequent
acquirer subsamples, the results are, however, inconsistent across different samples.
Given that some of the managerial talents are important in some of the regressions, we
examine further the addition to the explanatory power as a result of including these variables,
and whether they reduce the statistical and economic magnitude of the acquirer fixed effects.
Results are report in Panels B and C of Table 7. Due to different sample composition, the
acquirer fixed effects here are not directly comparable to those estimated in Table 2. We
therefore compute acquirer fixed effects for the three samples without managerial talent and
then repeat the estimation process by including managerial talent variables. We hold the
number of observations constant to facilitate comparisons. Essentially, the R2 (adjusted R2) do
not change when we include the managerial talent variables. Similarly, the inter-quartile
18
ranges reported in Panel C for the three models do not remarkably change with the inclusion
of managerial talent variables. In short, we find that the addition of the managerial talent
variables does not detract from the statistical and economic significance of the fixed effects,
suggesting that these characteristics are also orthogonal to the acquirer fixed effects
documented in this paper.
[Please Insert Table 7 Here]
Finally, investment bank advisors are unlikely to explain the fixed effects in acquirer
returns given that acquirers tend to switch advisors from deal to deal. Bao and Edmans (2011)
report that only 21.4% of deals in their sample are advised by the same investment bank as all
prior deals of the same acquirer in a 5-year period. In unreported results, we verify that the
inclusion of investment bank fixed effects does not detract from the economic and statistical
significance of acquirer fixed effects. We also note that the economic magnitude of the firm
specific effect, as measured by the inter-quartile range in the estimated acquirer fixed effects,
is several times larger than that of the advisor-specific effects documented by Bao and
Edmans (2011).
IV. Discussion
The results we document are consistent with acquirers possessing acquisition skill.
Jaffe, Pedersen and Voetmann (2013) analyze skill differences in acquisitions by regressing
acquirer returns in a given deal on the return of the same acquirer in its previous deal. They
show that such a positive dependence exists, but only when the two deals are conducted by
the same CEO (though noting that the latter finding may be due to low power arising from
very few CEO moves). Our approach is much more general in that our econometric
methodology allows us to study acquirer fixed effects in all deals by the same acquirer – be
they prior to or even after the deal in question. A finding of persistent acquirer returns is
broadly consistent with the findings of Kaplan and Schoar (2005) who document persistence
19
in the returns of private equity (buyout and venture capital) funds, whose business is
acquiring public and private firms. There also appears to be persistence in the performance of
serial entrepreneurs funded by venture capitalists (Gompers et al. (2010)).
Our results also add to the results of Fuller et al. (2002), Billett and Qian (2008) and
Aktas et al. (2011) who study serial acquirers and show that, for a given acquirer,
performance declines from deal to deal. While not contradicting those findings, in this paper
we show that some acquirers persistently perform above or below average and thereby
generate or destroy value by doing deals. To get a sense of the shape of the acquisition skill
distribution, Figure 1 Panel A presents the frequency chart of the estimated acquirer fixed
effects for the alternative serial acquirer sample, and Panel B does the same for the FNS
serial acquirer sample. The distribution is reasonably symmetric. If one interprets the
estimated fixed effects as firm-specific acquisition skill, there is a great degree of variation in
acquisition ability. While most of the mass is naturally around the mean, there are also many
extreme performers.
[Please Insert Figure 1 Here]
In order to shed more light on the extreme performers and their attributes, we examine
the identity of the best and the worst acquirers and demonstrate the fixed effects for these
individual firms.
[Please Insert Table 8 Here]
Table 8 provides information on the top 10 and bottom 10 acquirers sorted by acquirer
fixed effects for the two frequent acquirer subsamples (we do not report these results for the
full sample as most of the extreme performers have conducted only one deal, and their fixed
effects are not precisely estimated). Panel A reports the identity of the acquirer, the estimated
fixed effect, the average CAR (RET), the average residual CAR (RETRES) and the industry
affiliation for the alternative serial acquirer sample, and Panel B repeats this for the FNS
20
serial acquirer subsample. Naturally, negative (positive) fixed effects are associated with
negative (positive) CARs, though the relation is not monotonous given that the estimated
fixed effects are after controlling to firm and deal specific characteristics. For example, in the
FNS sample, AT&T has the fourth largest fixed effect of 13.83% with an average return of
1.16%, whilst Avant! Corporation has a fixed effect of 9.07% but an average CAR of 7.11%.
This pattern persists in the alternative serial acquirer subsample. Interesting observations
emerge when we consider the industry affiliation of the best and the worst acquirers.
Specifically, using Fama-French 48 industry groupings, we do not find a great deal of overlap
in the industry classifications of the top 10 and bottom 10 acquirers. This may suggest that
industry characteristics, which are subsumed by the firm fixed effects, may have a role to
play. In order to establish whether certain industries are systematically associated with
high/low fixed effects acquirers we regress the estimated fixed effects on a set of Fama-
French 48 industry dummies. We find that the industry dummies are jointly significant.
However, the R2 (Adj. R2) of these regressions are between 1.67% (0.53%), 3.70% (1.66%),
and 13.38% (2.52%) and for full, alternative, and FNS serial acquirer samples respectively,
meaning that only a small fraction of the variation in acquirer fixed effects can be explained
by industry affiliation.8 These results further underpin the idea that forces that are unique to
an organization, and not its environment or the top management team determine
extraordinary acquisition performance.
Taken together, our results have important implications for the M&A literature.
Specifically, our results imply that despite the seeming saturation of the M&A returns
literature with numerous studies, we are still far away from understanding the drivers of, and
the variation in, takeover gains. Did thirty years of empirical M&A research get it wrong?
We do not think this is the case. Existing studies on the determinants of takeover gains are
8 This conclusion is further substantiated by the fact the inclusion of industry fixed effects in the baseline
regressions of Table 1 has a very modest effect on the explanatory power of those models.
21
highly informative. It is just that the effects they document do not appear to be first-order
ones. For instance, Moeller et al. (2004) size effect in acquirer returns is attributed to greater
agency problems in large firms. While one should be able to detect this governance quality in
the cross-section of acquirer returns (as one does), it does not seem that these effects should
be dominating. Similarly, many of the deal characteristics do not appear to be the major
drivers of acquirer returns. For example, Travlos (1987) method of payment effect in acquirer
returns is attributed to the adverse selection of issuing equity, and, as such, is a manifestation
of acquirer stand-alone value re-setting. But normally we should not expect this stand-alone
value revelation to dwarf the value implications of the deal itself (particularly for frequent
acquirers, who regularly reveal private information about their value through payment
method choices). Again, this is not to say that the effects identified by prior literature are
unimportant. In fact, the size effect and the method of payment effect we pointed out here are
among the most robust determinants of acquirer returns across various studies. Moreover, in
the results reported in Appendix B (Table B.2) we verify that the effects of various
determinants of acquirer returns used in Table 1 continue to remain significant in a firm fixed
effects specification, where the identification is coming from within-firm variation in those
variables. This suggests that prior findings on these determinants of acquirer returns are not
simply capturing time-invariant firm-specific heterogeneity. Nevertheless, our message is that
we are missing a much bigger piece of the puzzle, and that it appears to be firm-specific. We
hope our findings will inspire further research in this direction.
V. Conclusion
In this paper we show that a large proportion of the variation in acquirer returns can be
explained by a firm-specific, time-invariant factor. In fact, the explanatory power of the
acquirer fixed effects overshadows that of many of the major firm and deal specific
22
characteristics combined. Economically, acquirers in the top (bottom) quintile of acquirer
fixed effects demonstrate performance that is an order of magnitude larger (smaller) than the
average acquirer return. We further show that acquirer returns are persistent over time.
Extraordinary good acquirers continue to make good acquisitions, while bad acquirers
continue to perform poorly.
We further examine the economic forces behind the statistical concept of acquirer fixed
effects. We investigate whether the fixed effects can be attributed to the characteristics of the
acquiring firm’s management team. However, various attributes of the managerial team
found to be important in the management science literature do not explain the fixed effect
away. Alternatively, the firm fixed effect could be picking up the CEO fixed effect. We are
unable to rule this explanation out due to lack of CEO moves between acquiring firms – a
data structure that would allow for separate identification of firm and CEO fixed effects.
Finally, there is a great deal of industry overlap between the best and the worst acquirer. We
conclude that acquisition skill (or lack thereof) is determined by the nature of the firms’
assets (tangible or intangible), or is hardwired into the organizational structure of some firms.
However, more interesting questions remain. For instance, are extraordinary acquirers born or
made? Further research in this direction could help shed more light on the sources of
persistent acquirer returns. A close examination of the best and worst acquirers that we
identify as part of our research design could serve as a potential starting point. To conclude,
despite the perceived saturation of the finance literature with M&A returns studies, the quest
for the determinants of takeover gains and their variation is far from over.
23
REFERENCES
Aktas, Nihat, Eric de Bodt, and Richard Roll, 2009, Learning, hubris and corporate serial acquisitions, Journal of Corporate Finance 15, 543-561.
Aktas, Nihat, Eric de Bodt, and Richard Roll, 2011, Serial acquirer bidding: An empirical test of the learning hypothesis, Journal of Corporate Finance 17, 18-32.
Aktas, Nihat, Eric de Bodt, and Richard Roll, 2013, Learning from repetitive acquisitions: Evidence from the time between deals, Journal of Financial Economics 108, 99-117.
Bao, Jack, and Alex Edmans, 2011, Do Investment Banks Matter for M&A Returns?, Review of Financial Studies 24, 2286-2315.
Bertrand, Marianne, and Antoinette Schoar, 2003, Managing with Style: The Effect of Managers on Firm Policies, Quarterly Journal of Economics 118, 1169-1208.
Billett, Matthew T., and Yiming Qian, 2008, Are Overconfident CEOs Born or Made? Evidence of Self-Attribution Bias from Frequent Acquirers, Management Science 54, 1037-1051.
Carhart, Mark M., 1997, On Persistence in Mutual Fund Performance, Journal of Finance 52, 57-82.
Chang, Yuk Ying, Sudipto Dasgupta, and Gilles Hilary, 2010, CEO Ability, Pay, and Firm Performance, Management Science 56, 1633-1652.
Chemmanur, Thomas J., and Imants Paeglis, 2005, Management quality, certification, and initial public offering, Journal of Financial Economics 76, 331-368.
Custódio, Cláudia, and Daniel Metzger, 2013, How Do CEOs Matter? The Effect of Industry Expertise on Acquisition Returns, Review of Financial Studies 26, 2008-2047.
Fee, C. Edward, Charles J. Hadlock, and Joshua R. Pierce, 2013, Managers with and without Style: Evidence Using Exogenous Variation, Review of Financial Studies 26, 567-601.
Fich, Eliezer M., Tu Nguyen, and Micah S. Officer, 2013, Large Wealth Creation in Mergers and Acquisitions, Working Paper.
Finkelstein, Sydney, 1992, Power in Top Management Teams: Dimensions, Measurement, and Validation, Academy of Management Journal 35, 505-538.
Finkelstein, Sydney, and Donald C. Hambrick, 1990, Top-Management-Team Tenure and Organizational Outcomes: The Moderating Role of Managerial Discretion, Administrative Science Quarterly 35, 484-503.
Fuller, Kathleen, Jeffry Netter, and Mike Stegemoller, 2002, What Do Returns to Acquiring Firms Tell Us? Evidence from Firms That Make Many Acquisitions, Journal of Finance 57, 1763-1793.
Golubov, Andrey, Dimitris Petmezas, and Nickolaos G. Travlos, 2012, When It Pays to Pay Your Investment Banker: New Evidence on the Role of Financial Advisors in M&As, Journal of Finance 67, 271-312.
Gompers, Paul, Anna Kovner, Josh Lerner, and David Scharfstein, 2010, Performance persistence in entrepreneurship, Journal of Financial Economics 96, 18-32.
Graham, John R., Si Li, and Jiaping Qiu, 2012, Managerial Attributes and Executive Compensation, Review of Financial Studies 25, 144-186.
Haleblian, Jerayr, and Sydney Finikelstein, 1993, Top Management Team Size, CEO Dominance, and Firm Performance: The Moderating Roles of Environmental Turbulence and Discretion, Academy of Management Journal 36, 844-863.
Hambrick, Donald C., Theresa Seung Cho, and Ming-Jer Chen, 1996, The Influence of Top Management Team Heterogeneity on Firms' Competitive Moves, Administrative Science Quarterly 41, 659-684.
24
Hambrick, Donald C., and Richard A. D'Aveni, 1992, Top Team Deterioration as Part of the Downward Spiral of Large Corporate Bankruptices, Management Science 38, 1445-1466.
Harford, Jarrad, Mark Humphery-Jenner, and Ronan Powell, 2012, The sources of value destruction in acquisitions by entrenched managers, Journal of Financial Economics 106, 247-261.
Hartzell, Jay C., Eli Ofek, and David Yermack, 2004, What's In It for Me? CEOs Whose Firms Are Acquired, Review of Financial Studies 17, 37-61.
Hayward, Mathew L. A., and Donald C. Hambrick, 1997, Explaining the Premiums Paid for Large Acquisitions: Evidence of CEO Hubris, Administrative Science Quarterly 42, 103-127.
Jaffe, Jeffrey, David Pedersen, and Torben Voetmann, 2013, Skill Differences in Corporate Acquisitions, Journal of Corporate Finance forthcoming.
Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency, Journal of Finance 48, 65-91.
Jehn, Karen A., 1995, A Multimethod Examination of the Benefits and Detriments of Intragroup Conflict, Administrative Science Quarterly 40, 256-282.
Jensen, Michael C., and Richard S. Ruback, 1983, The market for corporate control, Journal of Financial Economics 11, 5-50.
Kaplan, Steven N., and Antoinette Schoar, 2005, Private Equity Performance: Returns, Persistence, and Capital Flows, Journal of Finance 60, 1791-1823.
Lemmon, Michael L., Michael R. Roberts, and Jaime F. Zender, 2008, Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure, Journal of Finance 63, 1575-1608.
Masulis, Ronald W., Cong Wang, and Fei Xie, 2007, Corporate governance and acquirer returns, Journal of Finance 62, 1851-1889.
Miller, C. Chet, Linda M. Burke, and William H. Glick, 1998, Cognitive diversity among upper-echelon executives: Implications for strategic decision processes, Strategic Management Journal 19, 39.
Moeller, Sara B., Frederik P. Schlingemann, and René M. Stulz, 2004, Firm size and the gains from acquisitions, Journal of Financial Economics 73, 201-228.
Moeller, Sara B., Frederik P. Schlingemann, and René M. Stulz, 2005, Wealth destruction on a massive scale? A study of acquiring-firm returns in the recent merger wave, Journal of Finance 60, 757-782.
Roll, Richard, 1986, The Hubris Hypothesis of Corporate Takeovers, Journal of Business 59, 197-216.
Travlos, Nickolaos G., 1987, Corporate takeover bids, methods of payment, and bidding firms' stock returns, Journal of Finance 42, 943-963.
25
Figure 1Distribution of Acquirer Fixed Effects
The figures depict the frequency distribution of the estimated acquirer fixed effect for the alternative serial acquirer (Panel A) and the FNS serial acquirer (Panel B) samples. The graphs are drawn using histograms and the kernel density estimation (curved line). Acquirer fixed effects are estimated using the regression model (4) of Table 2. Similar to Graham et al. (2012), the fixed effects are normalized so that the mean value is zero. This does not alter the shape of the distribution and its variance. In Panel A, an outlier (eMedSoft.com with a fixed effect of 76.45%) has been removed.
This table presents the results of OLS regressions of acquirer CARs on acquirer and deal characteristics for the full sample as well as the two subsamples of serial acquirers. The full sample includes all domestic M&A transactions completed during the period 1990-2011 from the Thomson Financial SDC M&A Database. The subsamples are classified based on the two alternative definitions of serial acquirers. "Alternative" serial acquirers are defined as those having completed two or more deals over a 3-year window. FNS (Fuller, Netter and Stegemoller (2002)) serial acquirers are defined as those having completed at least five deals over a 3-year window. The dependent variable in all the specifications is the cumulative abnormal returns of the acquiring firm stock over the event window (-2, +2) surrounding the announcement date. The return is based on the market model with the benchmark being the CRSP value-weighted index. The t-statistics in parentheses are adjusted for heteroskedasticity. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively. The variables are defined in Appendix A. Full Sample Alternative Serial FNS Serial
This table reports the joint significance of acquirer fixed effects in the regression model of acquirer CARs for the full sample (Panel A) and the two definitions of serial acquirers (Panels B and C). Acquirer CARs are regressed on acquirer fixed effects and the control variables specified in models (1) to (4). Deal characteristics include relative size, relatedness, tender, and hostile indicators, and full set of target listing status/payment method interactions. Acquirer characteristics include the natural logarithm of acquirer size, Tobin’s Q, free cash flow, leverage, run-up and sigma. F-statistics for the joint significance of acquirer fixed effects are reported, along with their corresponding p-values and the number of firms (in parentheses). The R2 and the Adjusted R2 of the models are also shown. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.Row Controls Acquirer FE F-test N R2 Adj. R2
Panel A: Full sample(1) None 1.728*** (0.000,4128) 12491 0.460 0.194(2) Year FE 1.729*** (0.000,4128) 12491 0.464 0.197(3) Deal chars., year FE 1.725*** (0.000,4128) 12491 0.478 0.217(4) Acquirer and deal chars., year FE 1.690*** (0.000,4128) 12491 0.487 0.231Panel B: Alternative serial acquirers(1) None 1.368*** (0.000,2219) 9373 0.298 0.080(2) Year FE 1.363*** (0.000,2219) 9373 0.302 0.083(3) Deal chars., year FE 1.273*** (0.000,2219) 9373 0.319 0.104(4) Acquirer and deal chars., year FE 1.280*** (0.000,2219) 9373 0.331 0.119Panel C: FNS serial acquirers(1) None 1.211*** (0.009,333) 2611 0.150 0.026(2) Year FE 1.232*** (0.005,333) 2611 0.165 0.035(3) Deal chars., year FE 1.102 (0.116,333) 2611 0.178 0.046(4) Acquirer and deal chars., year FE 1.242*** (0.003,333) 2611 0.197 0.066
28
Table 3Distribution of Acquirer Fixed Effects
Panel A describes the distribution of the estimated acquirer fixed effects. Panel A presents the standard deviation, the 25th percentile, the 75th percentile, and the inter-quartile range of the estimated fixed effects in the full sample. Panels B and C repeat the same statistics for the alternative serial acquirers and the FNS serial acquirers samples, respectively.Panel A: Full sample
This table presents the results of a placebo test of acquirer fixed effects. The estimated models are identical to those in Table 2, but the data are scrambled such that deals are randomly allocated to firms. We run these Monte Carlo permutations 1000 times and report the median value of the F-test for the joint significance of the estimated firm fixed effects and the associated p-values. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively. Controls Acquirer FE F-test p-valuePanel A: Full sample (1) None 0.991 0.624(2) Year FE 0.992 0.618(3) Deal chars., year FE 0.998 0.533(4) Acquirer and deal chars., year FE 0.997 0.544Panel B: Alternative serial acquirers (1) None 0.998 0.519(2) Year FE 0.998 0.524(3) Deal chars, year FE 0.997 0.535(4) Acquirer and deal chars., year FE 0.998 0.525Panel C: FNS serial acquirers (1) None 0.998 0.500(2) Year FE 0.998 0.502(3) Deal chars., year FE 0.999 0.499(4) Acquirer and deal chars., year FE 0.998 0.501
30
Table 5Persistence of Acquirer Returns
This table presents univariate tests of persistence in acquirer returns for the two definitions of serial acquirers. In Panel A serial acquirers are sorted into quintiles based on their average CARs (denoted RET) over the last 3 calendar years. Q1 and Q5 represent serial acquirers with the lowest and highest past RET, respectively. The average CARs to acquisitions made by all the acquirers in Q1 and Q5 over the next k calendar years are then computed, where k = (1, 2, 3, 4, 5) and denoted as future RET. Panel B repeats the analysis where residual CARs obtained from regressions estimated in Table 1 are used to sort acquirers into performance quintiles (past RETRES) and to measures subsequent performance (future RETRES). Panels C and D examine persistence in transaction-value weighted RET and RETRES, respectively, where the weights are the ratios of the deal value to the sum of deals values of the given acquirer over a period in which the performance is measured. The t-statistics for the differences in means between Q5 and Q1 are reported in parentheses. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.Panel A: Persistence in CARs Future RET measured overQuintiles measured over 3yr RET 1yr 2yr 3yr 4yr 5yrAlternative serial acquirersQ1 0.41% -0.17% -0.17% -0.15% -0.15%Q5 0.72% 0.90% 0.91% 1.16% 1.22%Q5-Q1 0.31% 1.07% 1.08% 1.31% 1.37%
Table 6Regression Analysis of Persistence in Acquirer Returns
This table presents the results of OLS and quantile regressions of future returns on past returns for the two serial acquirer subsamples. Panels A, B, and C estimate the mean (OLS), 20th percentile, and 80th percentile of the future returns distribution, respectively. The dependent variable is RET measured as the average CARs to all the acquisitions made by an acquirer over the next k calendar years, where k = (1, 2, 3, 4, 5). The explanatory variable ‘Past RET’ is the average CAR to all acquisitions over the last 3 calendar years. For the OLS regressions the t-statistics in parentheses are adjusted for clustering by acquirer. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.Panel A: OLS regressions
Future RET measured over 1yr 2yr 3yr 4yr 5yrAlternative Serial SampleIntercept -0.0012 -0.0005 0.0008 0.001 0.0014
Table 7Managerial Talent, Acquirer CARs and Acquirer Fixed Effects
Estimates reported in Panel A are from regressions of acquirer CARs on acquirer top management team characteristics and other controls listed in Table 1 for the full sample as well as the two subsamples of serial acquirers. Only the coefficients on the top management team variables are reported, with the t-statistics in parentheses adjusted for heteroskedasticity. Panel B reports the joint significance of acquirer fixed effects in the regression model of acquirer CARs on acquirer fixed effects, year fixed effects, the deal and acquirer characteristics, and with and without the top management team variables. F-statistics for the joint significance of acquirer fixed effects are reported, along with their corresponding p-values and number of firms (in parentheses). The R2 and the Adjusted R2 of the models are also shown. Panel C reports the distribution of acquirer fixed effects for the full sample as well as for the subsamples of serial acquirers before and after the top management team variables are added. Team size is measured as the number of the acquiring firm's officers with a listed title above the vice president level in the most recent fiscal year prior to the announcement date. Average tenure represents the average number of years for which the top management team members have worked in the acquiring firm prior to the announcement date. Tenure heterogeneity is the coefficient of variation of the top management team members’ tenures. The average age of top management team members is measured at the end of the most recent fiscal year prior to the announcement date. CEO dominance is calculated as CEO’s salary and bonus divided by the average salary and bonus of other team members for the most recent fiscal year prior to the announcement date. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively. Panel A: OLS regression of CARs on Managerial Talent variables All Alternative FNS
Team Size 0.0014 0.0011 -0.0013(0.9637) (0.6726) (-0.3973)
Average Tenure 0.0003 0.0005** -0.0001(1.6092) (2.0604) (-0.2055)
N 2188 1799 576R2 (Adj. R2) 0.085 (0.068) 0.097 (0.077) 0.122 (0.062)Panel B: Managerial Talent and Acquirer FE Without Managerial Talent With Managerial Talent Acq. FE F-test R2 Adj.
R2 Acq. FE F-test R2 Adj.
R2N
(1) Full 1.133** (0.024,760) 0.432
0.110
1.129**(0.028,760)
0.433 0.109 2188
(2) Alternative
1.145** (0.032,509) 0.380
0.113
1.143**(0.034,509)
0.383 0.114 1799
(3) FNS 1.473***(0.005,96) 0.313
0.118
1.324**(0.033,96) 0.316 0.112 576
Panel C: Distribution of Acquirer FE with Managerial Talent Controls With Managerial Talent SD 25th 75th Inter-quar.
Rng. SD 25th 75th Inter-quar.
Rng.(1) Full
5.83%-
2.62%2.82% 5.44%
5.89%
-2.81%
2.85% 5.66%
(2) Alternative 5.06%
-2.10%
2.81% 4.91%
5.15%
-2.44%
2.83% 5.27%
35
(3) FNS3.88%
-2.30%
2.75% 5.05%
3.97%
-2.79%
3.20% 5.99%
36
Table 8The Best and the Worst Acquirers
This table reports the top and bottom 10 acquirers ranked by acquirer fixed effects estimated in model (4) of Table 3 for the two alternatively definitions of serial acquirers. RET (RETRES) is measured as the average CARs (residuals) over the full sample period for individual acquirers.Panel A: The top and bottom 10 acquirers in the alternative serial acquirers sample
Acquirer FE RETRETRE
S N FF48Top 10 Acquirers e-MedSoft.com 76.45% 83.26% 75.59% 2 Business ServicesMet Capital Corp 41.00% 48.88% 42.65% 2 TradingE For M Corp 39.11% 45.73% 40.29% 2 RecreationPolyVision Corp 36.25% 21.72% 12.49% 2 Meas. & Control Equip.Ask Jeeves Inc 32.71% 36.72% 33.82% 2 Business ServicesCatalytica Inc 30.37% 36.49% 31.13% 2 ChemicalsCapitol Multimedia Inc 27.23% 36.35% 30.64% 3 EntertainmentStar Technologies Inc 27.00% 37.92% 30.22% 2 ComputersEvans Inc 24.93% 34.34% 27.56% 2 RetailSunRiver Corp 24.10% 32.30% 25.39% 2 ComputersBottom 10 Acquirers
GlobeSpan Inc -30.88%-
17.72% -32.09% 2 Electronic Equipment
OrthoLogic Corp -29.28%-
14.89% -17.30% 2 Pharmaceutical Products
Socrates Technologies Corp -28.09%-
20.36% -25.51% 2 Meas. & Control Equip.
L90 Inc -23.50%-
15.94% -20.33% 2 Business Services
Trinzic Corp -20.25%-
14.31% -19.16% 2 Business ServicesPrice Communications Corp -20.17% -7.65% -21.69% 2 Communication
Deltagen Inc -19.57%-
16.50% -18.24% 2 Pharmaceutical Products
Gasco Energy Inc -19.43%-
17.24% -17.92% 2 Petroleum and Nat. Gas
ARC Capital -19.26%-
14.47% -17.89% 2 Meas. & Control Equip.
Pre-Paid Legal Services Inc -18.68%-
19.30% -18.92% 2 Personal ServicesTable continues on the next page
Table 7 - continuedPanel B: The top and bottom 10 acquirers in the FNS serial acquirers sample
Acquirer FE RETRETRE
S N FF48Top 10 Acquirers Veeco Instruments Inc 17.35% 13.69% 12.96% 5 Meas. & Control Equip.PMC-Sierra Inc 14.44% 3.07% 3.38% 6 Electronic EquipmentMedtronic Inc 13.88% 0.63% 3.10% 5 Medical EquipmentAT&T Corp 13.83% 1.16% 5.10% 5 CommunicationChattem Inc 13.51% 17.24% 15.41% 5 Pharmaceutical ProductsEnvirofil Inc 12.49% 9.67% 10.38% 5 Business ServicesGuidant Corp 9.75% 1.84% 3.19% 5 Medical EquipmentAvant! Corp 9.07% 7.11% 7.06% 6 Business ServicesCisco Systems Inc 8.69% -1.05% -0.23% 6 ComputersComerica Inc 8.35% 0.89% 2.31% 6 BankingBottom 10 Acquirers Digital River Inc -13.94% -4.31% -4.67% 5 Business ServicesQuestron Technology Inc -13.70% -3.42% -7.87% 9 Business Services
37
Ultralife Corp -11.34% -4.21% -5.81% 5 Electrical EquipmentKratos Defense & Security -11.25% -4.76% -5.43% 8 CommunicationInterland Inc -10.40% -4.96% -5.31% 7 ComputersEltrax Systems Inc -10.36% -3.90% -5.52% 16 ComputersAdCare Health Systems Inc -10.20% 0.89% -1.00% 6 Healthcare24/7 Media Inc -9.00% -7.54% -7.29% 5 Business ServicesOmega Health Systems Inc -8.99% -0.44% -2.88% 7 HealthcareTBA Entertainment Corp -8.86% -1.24% -3.62% 7 Rests., Hotels, Motels
Appendix A
Variable Definitions
Variable DefinitionPanel A: Return Variables CAR (-2, +2) Cumulative abnormal return of the acquiring firm stock over the event
window (-2, +2) surrounding the announcement date. The return is calculated using the market model with the benchmark being the CRSP value-weighted index. The model parameters are estimated over the (-300, -91) period prior to the announcement.
Future RET Average CAR (-2, +2) to all the acquisitions made by an acquirer over the next k calendar years, where k = (1, 2, 3, 4, 5).
Past RET Average CAR (-2, +2) to all the acquisitions made by an acquirer over the last 3 calendar years.
RETRES Average residual from an OLS regression of CAR specified in Table 1.Panel B: Acquirer Characteristics
Acquirer Size The market value of the acquiring firm’s equity 11 days before the announcement date in $US dollar million. The data is obtained from CRSP.
Tobin’s Q Market value of the acquiring firm’s assets divided by book value of its assets for the fiscal year prior to the acquisition. The market value of assets is equal to book value of assets plus market value of common stock minus book value of common stock minus balance sheet deferred taxes. The data is obtained from both CRSP and Compustat.
Leverage The sum of the acquiring firm’s long-term debt and short-term debt divided by the market value of its total assets measured at the end of the fiscal year prior to the acquisition. The data is obtained from both CRSP and Compustat.
Free Cash Flow The acquiring firm’s operating income before depreciation minus interest expense minus income tax plus changes in deferred taxes and investment tax credits minus dividends on both preferred and common share divided by its book value of total assets at the fiscal year-end before the announcement date from Computstat.
Sigma Standard deviation of the market-adjusted daily returns of the acquirer’s stock over a 200-day window (-210, -11) from CRSP.
Run-up Market-adjusted buy-and-hold return of the acquirer’s stock over a 200-day window (-210, -11) from CRSP.
Panel C: Deal Characteristics Public Indicator variable: one if the bid is for a public target and zero otherwise.Private Indicator variable: one if the bid is for a private target and zero otherwise.Subsidiary Indicator variable: one if the bid is for a subsidiary target and zero otherwise.All Cash Indicator variable: one if the payment is pure cash and zero otherwise.Stock Indicator variable: one if the payment includes stock and zero otherwise.Relative Size The deal value from Thomson Financial SDC divided by the market value of
38
the bidding firm’s equity 11 days prior to the announcement date from CRSP.
Relatedness Indicator variable: one if the bidder and the target are operating in the same industries with a common 2-digit SIC code and zero otherwise. Data from Thomson Financial SDC.
Hostile Indicator variable: one if the deal is classified as ‘hostile’ by Thomson Financial SDC and zero otherwise.
Tender Offer Indicator variable: one if the deal is a tender offer and zero otherwise. Data from Thomson Financial SDC.
Panel D: Managerial Variables New CEO Indicator variable: one if an acquirer has employed a new CEO over the
period in which future RET is measured, and zero otherwise.Team Size The size of the acquiring firm’s top management team. It equals the number
of officers with a listed title above the vice president level in the most recent fiscal year prior to the announcement date. Data from ExecuComp.
CEO Dominance It is computed as CEO salary and bonus divided by the average salary and bonus of other team members for the most recent fiscal year prior to the announcement date. Data from ExecuComp
Average Tenure The average number of years for which top team members have worked in the acquiring firm prior to the announcement date. Data from ExecuComp
Tenure Heterogeneity The coefficient of variation of the top team members’ tenures.Average Age The average age of the top team members measured at the end of the most
recent fiscal year prior to the announcement date. Data from ExecuComp
39
Appendix B
Additional Results
Table B.1Regression Analysis of Persistence in Residual Returns
This table presents the results of OLS and quantile regressions of future residual returns on past residual returns for the two serial acquirer subsamples. Panels A, B, and C estimate the mean (OLS), 20th percentile, and 80th percentile of the future residual returns distribution, respectively. The dependent variable is RETRES measured as the average residual CARs to all the acquisitions made by an acquirer over the next k calendar years, where k = (1, 2, 3, 4, 5). The explanatory variable ‘Past RETRES’ is the average residual CAR to all acquisitions over the last 3 calendar years. For the OLS regressions the t-statistics in parentheses are adjusted for clustering by acquirer. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively. Panel A: OLS regressions
Future RETRES measured over 1yr 2yr 3yr 4yr 5yrAlternative Serial SampleIntercept -0.0033 -0.0029 -0.0018 -0.0008 -0.0001
Table B.2Determinants of Acquirer CARs - Using Within-Firm Variation
This table presents the results of OLS regressions of acquirer CARs on acquirer and deal characteristics, for the full sample as well as the two subsamples of serial acquirers controlling for firm fixed effects. The full sample includes all domestic M&A transactions completed during the period 1990-2011 from the Thomson Financial SDC M&A Database. The subsamples are classified based on the two alternative definitions of serial acquirers. "Alternative" serial acquirers are defined as those having completed two or more deals over a 3-year window. FNS (Fuller, Netter and Stegemoller (2002)) serial acquirers are defined as those having completed at least five deals over a 3-year window. The dependent variable in all the specifications is the cumulative abnormal returns of the acquiring firm stock over the event window (-2, +2) surrounding the announcement date. The return is based on the market model with the benchmark being the CRSP value-weighted index. The t-statistics in parentheses are adjusted for heteroskedasticity. Symbols ***, ** and * denote significance at the 1%, 5% and 10% level, respectively. The variables are defined in the Appendix. All Alternative FNS
